Which LoRAs work here?

Files trained for LTX-2.3 (the 22B model), such as the style, effect and character LoRAs listed under LTX-2.3 on Hugging Face and Civitai. LoRAs for LTX-2 19B, Wan, HunyuanVideo or image models will not give you their look on this model.

Can I use a Civitai link?

Yes. The civitai.com/api/download/models/... link of a LoRA loaded in our test and gave the same clip as the Hugging Face copy of the same file. Use the download link, not the model page.

Why did my LoRA fail?

The three causes we saw: the file is over 2 GiB, the Hugging Face file needs a login (most of Lightricks' own LTX-2.3 LoRAs do), or the link points to a repository page instead of the file. None of these is charged.

My LoRA loaded but nothing changed. Why?

It was most likely trained for a different base model: in our test a Wan LoRA loaded without an error and the clip matched the run without it. Check the trigger words too, and on image-to-video try a higher strength.

What strength should I use?

On text-to-video, 0.5 to 1 for styles and effects. On image-to-video start at 1 to 1.3: the opening photo holds its look, and above about 1.3 a style LoRA can change faces.

How is it different from LTX 2.3 Spicy LoRA?

Here you bring any LoRA file by link, on text-to-video and image-to-video. LTX 2.3 Spicy LoRA is a separately tuned build whose loras field only re-weights its own built-in LoRAs.

Is LTX-2.3 LoRA uncensored for NSFW video?

Yes. Your LoRAs are not reviewed and the output is not filtered. Without a LoRA the model rendered the milder levels of our content test as written and held part of the scene back at the strongest levels; with community LoRAs trained for the stronger levels it went further in our tests. Every person depicted must be an adult.

Per second of output at the chosen resolution, plus a fixed fee for each clip. Loading one, two or three LoRAs costs the same, the shape does not change the price, and failed requests are not charged. The live estimate before you submit shows the full amount.

LTX-2.3 LoRA API: Uncensored NSFW Video, Your LoRAs

lightricks/ltx-2.3-lora/text-to-video

LTX-2.3 LoRA text-to-video loads up to three of your own LoRA files on Lightricks' 22B audio-video model and returns a 5 to 20 second clip with stereo sound at 480p, 720p or 1080p. Styles, effects and characters by Hugging Face or Civitai link, no content filter.

From $0.024/s

Run model View API

Input

Text to video

Prompt

Aspect ratio

Resolutionaffects price

Durationaffects price

Style LoRAs0/3

Seed

Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run

Example result

Extend

Video upscaling

Video editing

{
  "model": "lightricks/ltx-2.3-lora/text-to-video",
  "input": {
    "aspect_ratio": "16:9",
    "resolution": "720p",
    "duration_seconds": 5,
    "seed": 4111,
    "prompt": "A woman in her late twenties with short natural curls and warm brown skin, wearing a cobalt-blue satin slip dress, walks slowly away from the camera down an empty glass aquarium tunnel after closing time, barefoot, her heels dangling from one hand. A giant manta ray glides overhead and its shadow sweeps across her; blue caustic light ripples over her bare shoulders and back. She stops, turns her head and gives a smouldering glance back over her shoulder. The camera follows behind her, then holds. 35mm anamorphic, deep teal and cobalt grade, film grain. Audio: muffled hum of water, soft bare footsteps, one low cello note."
  }
}

Introduction

LTX-2.3 LoRA Text to video generates a clip with its own soundtrack from a prompt, with up to three of your LoRA files loaded. It runs Lightricks' open 22-billion-parameter audio-video model; clips run 5 to 20 seconds at 480p, 720p or 1080p in 16:9 or 9:16. The platform adds no content filter to the prompt, the weights or the clip.

Choose a mode by starting material

What you have	Mode
a written description	LTX-2.3 LoRA Text to video
a photo to animate	LTX-2.3 LoRA Image to video
a LoRA made for LTX-2 19B	LTX-2 19B LoRA Text to video

Key features & highlights

Style LoRAs that take over the frame

In our tests a claymation LoRA and an oil-painting LoRA each restyled the whole scene; without the files the same prompts stayed photographic.

Links from Hugging Face or Civitai

A direct Hugging Face file link and a Civitai download link both loaded, and gave the same clip for the same file.

Sound on every clip

Each clip has a stereo soundtrack generated with the picture; describe it on an "Audio:" line.

Uncensored with custom weights

LoRAs are not reviewed and the output is not filtered.

Request parameters

Request parameters for LTX-2.3 LoRA · Text to Video
Parameter	Required	Description
`prompt`	Yes	Trigger words first if your LoRA has them, then the shot, the action and the camera, then an "Audio:" line for the soundtrack. at least 1 characters
`aspect_ratio`	No	16:9 for landscape (the default) or 9:16 for portrait. The shape does not change the price. `16:9` · `9:16` · Default: `16:9`
`resolution`	No	480p (832 x 448), 720p (1280 x 704, the default) or 1080p (1920 x 1024). Each step costs more per second. `480p` · `720p` · `1080p` · Default: `720p`
`duration_seconds`	No	Any whole second from 5 to 20, billed per second on top of the fixed per-clip fee. `≥ 5` · `≤ 20` · Default: `5`
`loras`	No	Up to three {path, scale} entries trained for LTX-2.3; each file under 2 GiB, scale 0 to 4 (0.5 to 1 is a good start for styles and effects). Default: `[]` at most 3 items
`seed`	No	Keeps runs comparable while you change one LoRA strength at a time. It does not promise a frame-identical clip. `≥ 0` · `≤ 2147483647`

Pricing

Billed per second of output at the chosen resolution, plus a fixed fee for each clip; the live estimate shows the full amount. Loading LoRAs adds no further charge, and the aspect ratio does not change the price.

Pricing
Specification	Price
`480p`	$0.024 / second
`720p`	$0.036 / second
`1080p`	$0.048 / second

Use cases

Styled series

One style LoRA, many prompts, one consistent look.

Effect shots

Community effect LoRAs showed at strength 0.5 in our tests.

LoRAs your audience asks for

Load the LoRA your audience expects; adults only on screen.

Explore key use-case tutorials

Prompt guide & pro tips

Use LoRAs made for LTX-2.3

A LoRA trained for LTX-2 19B changed our clip unpredictably instead of applying its look; a Wan LoRA did nothing.

Compare against no LoRA

Style words alone already move the base model, so run the same prompt and seed with loras empty before judging a file.

Mind the 2 GiB limit

Larger files fail the request, which is not charged.

Use direct, public links

Link the .safetensors file or the Civitai download, not a repository page; files behind a login fail.

API quick start

# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SPICY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
  -d '{
  "model": "lightricks/ltx-2.3-lora/text-to-video",
  "input": {
    "aspect_ratio": "16:9",
    "duration_seconds": 5,
    "prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
    "resolution": "720p"
  }
}'

# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
  -H "Authorization: Bearer $SPICY_API_KEY"

See the API reference for authentication, the complete schema, polling, and error handling.

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

LORA VIDEO API

LTX-2.3 LoRA API at a glance

Lightricks' LTX-2.3 with your own LoRA files: styles, effects and characters by link, sound on every clip, no content filter.

LoRAs per call: 3; Up to three weight files per clip, each with its own strength from 0 to 4.
LoRA source: Link; A direct Hugging Face file link or a Civitai download link; both loaded in our tests.
File size limit: 2 GiB; Each LoRA file must be under 2 GiB; larger files fail the request without charge.
Sound: Stereo; Every clip has an audio track generated with the picture; describe it on an Audio: line.
Clip length: 5 – 20 s; Any whole second from 5 to 20, at 480p, 720p or 1080p.
Content filter: None; Uncensored: your LoRAs are not reviewed and the result is not filtered.

Results

What you can make with LTX-2.3 LoRA

Style LoRAs
One file, a whole new look
A community oil-painting LoRA at strength 1 turns an aquarium scene into a moving painting, brush strokes and all. The same prompt and seed without the file came back as an ordinary photographic clip.
Text-to-video with a community oil-painting LoRA at strength 1, 720p, five seconds.
Image to video
Restyle a photo as it moves
A photo of a man lying under a projected eclipse plus a claymation LoRA at 1.3: within the first second he becomes a sculpted clay figure who reaches for the camera, with clay stars falling around him.
Image-to-video with a community claymation LoRA at strength 1.3, 720p, five seconds.
Sound included
Picture and sound in one pass
Every clip comes back with a stereo soundtrack generated with the picture. Put the sound on its own "Audio:" line, as this near-kiss under a planetarium sky does.
Text-to-video without a LoRA, 720p, five seconds.

More clips made with LTX-2.3 LoRA

Text-to-video without a LoRA, 720p, five seconds.
5s
Image-to-video without a LoRA from a still, 720p, five seconds.
5s
Image-to-video without a LoRA from a still, 720p, five seconds.
5s
Image-to-video with a community oil-painting LoRA at strength 1.2, 720p, five seconds.
5s
Text-to-video in portrait with a community claymation LoRA at strength 1, 720p, five seconds.
5s

Content test

Is LTX-2.3 LoRA uncensored? What we tested

We ran each level on our live setup and checked every clip frame by frame. A pass means the content was actually on screen, not just that the request went through.

Content levels tested on LTX-2.3 LoRA, by mode
Level	Image to video	Text to video
MildThe lightest prompts in the test	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026
ModerateWhere filtered models start to refuse	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026
StrongThe middle of the scale	Not tested yet	RendersTested Oct 9, 2026
Very strongWhere most models start to hold back	Not tested yet	Not tested yet
MaximumThe top of the scale	RendersTested Oct 9, 2026	Not tested yet

Renders: on screen as written. Softened: generated, but milder than what was written. Refused: rejected. Not tested yet: we have not run that level on that mode.

How each mode behaves

All modes: Weights must be trained for LTX-2.3 and be under 2 GiB. A LoRA for another base model loads without an error and either changes nothing (a Wan LoRA in our test) or changes the clip unpredictably (an LTX-2 19B LoRA). Most of Lightricks' own LTX-2.3 LoRAs on Hugging Face sit behind a login, so they cannot be loaded by link, and IC-LoRAs that expect a reference video have no input for it here. Three LoRAs is a hard cap. Every clip comes with a soundtrack and there is no switch to turn it off. Without a LoRA the model rendered our content probes up to the middle levels as written and held part of the scene back at the strongest levels.
Text to video: Only two shapes: 16:9 and 9:16.
Image to video: The clip keeps the shape of your photo and a prompt is required. A style LoRA redraws your photo, so faces and details change; above strength 1.3 it can change who the person looks like.

Use cases

What to build with the LTX-2.3 LoRA API

A house style for a series
Load one style LoRA and every clip in the set shares the look: in our tests a claymation LoRA and an oil-painting LoRA each restyled the whole scene, set and people.
Restyle a photo as it moves
Image-to-video with a style LoRA turns your still into that style during the first second and animates it, with sound.
Effects and characters from the community
In our tests an effect LoRA showed at strength 0.5 on text-to-video and at 1 on image-to-video, so the large LTX-2.3 LoRA catalogues on Hugging Face and Civitai are usable as they are.
Uncensored community LoRAs
Character and style LoRAs of any kind load like any other file: no allow-list, no review, no filter on the result. Everyone on screen must be an adult.

Prompts

LTX-2.3 LoRA prompt examples you can copy

Paste one into the playground as it is, then change one thing at a time.

Text to video
A glance in the manta tunnel
Result from this prompt
A woman in her late twenties with short natural curls and warm brown skin, wearing a cobalt-blue satin slip dress, walks slowly away from the camera down an empty glass aquarium tunnel after closing time, barefoot, her heels dangling from one hand. A giant manta ray glides overhead and its shadow sweeps across her; blue caustic light ripples over her bare shoulders and back. She stops, turns her head and gives a smouldering glance back over her shoulder. The camera follows behind her, then holds. 35mm anamorphic, deep teal and cobalt grade, film grain. Audio: muffled hum of water, soft bare footsteps, one low cello note.
Why it works. No LoRA. The prompt says the camera follows behind her and then holds, so the clip ends on her glance back rather than a close-up.
Try this prompt
Image to video
A gown in the projector light
Result from this prompt
The woman in the backless champagne gown runs her fingertips along the brass star projector as it begins to rotate; the projected stars slide across her bare back and shoulders. She turns slowly toward the camera with an alluring glance and lifts her chin toward the dome as a bright comet streaks overhead. Slow orbit around her. Audio: the soft whir of the projector motor, a distant choir pad.
Why it works. No LoRA. The prompt names the light source, the projector starting to turn, so the stars slide across her back as she turns.
This mode starts from your own image: add it in the playground before you run.
Try this prompt
Text to video
A whisper in the planetarium seats
Result from this prompt
Close two-shot in a dark planetarium at night: a South Asian man in his late thirties in an unbuttoned white linen shirt and an East Asian woman in her early thirties in an off-the-shoulder black velvet dress recline side by side in tilted seats, the projected Milky Way slowly wheeling across the dome behind them. She leans over the armrest and whispers something in his ear; he laughs softly, turns his face to hers and their lips stop an inch apart as a shooting star streaks across the dome. Slow push-in. Indigo and silver starlight palette, soft haze, 50mm. Audio: a quiet ambient synth pad, her whisper, his low laugh.
Why it works. No LoRA. A close two-shot keeps both faces in frame for the near-kiss; the whisper and the laugh are on the Audio: line.
Try this prompt

Pricing

LTX-2.3 LoRA API pricing

A 5-second clip starts at $0.12 at 480p and $0.24 at 1080p.

The playground shows the exact price for your settings before you run. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.

See every price tier

Get started

How to use the LTX-2.3 LoRA API in three steps

Step 1
Pick a LoRA trained for LTX-2.3
Search Hugging Face or Civitai for LoRAs listed under LTX-2.3 (the 22B model). LoRAs made for LTX-2 19B load without an error here but do not reproduce their look: in our test one changed the clip in an unpredictable way instead of applying its style.
Step 2
Link the file, not the page
On Hugging Face copy the download link of the .safetensors file (it contains /resolve/main/); on Civitai the civitai.com/api/download/models/... link works. A repository page, a file behind a login or a file over 2 GiB fails, and is not charged.
Step 3
Trigger words first, then the shot
Start the prompt with the LoRA's trigger words, then who is in frame, what they do over the clip and what the camera does, and end with an "Audio:" line. Style words alone already nudge the base model, so compare against a run without the file before you judge a LoRA.

Open the playground Get an API key

Parameter reference and code samples are in the API and Readme tabs at the top of this page.

Why SpicyAPI

Why run LTX-2.3 LoRA on SpicyAPI

No filter from us
We add no content filter of our own. What the model renders is what you get back.
Tested, not guessed
Content labels come from real generations we ran and checked, with the test date on the page.
Pay per result, in dollars
No subscription and no credits. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.
One key for every model
One API key and balance cover image, video, audio and chat models, so you can explore more models from the same account.

FAQ

LTX-2.3 LoRA API FAQ

About LTX-2.3 LoRA

LTX-2.3 LoRA is Lightricks' open 22B audio-video model with room for your own weights: up to three LoRA files under 2 GiB per request, each with a strength from 0 to 4, linked from Hugging Face or Civitai, on text-to-video and image-to-video. Clips run 5 to 20 seconds at 480p, 720p or 1080p with a stereo soundtrack. LoRAs are not reviewed and nothing is filtered.

All LTX models6

LTX 2.5Text to video · Image to videoFrom $0.09per output second
LTX 2.3 SpicyImage to video · Video extensionSpicy list price: $0.02From $0.019per output second
LTX 2.3 Spicy LoRAImage to video · Video extensionSpicy list price: $0.03From $0.029per output second
LTX-2.3 LoRACurrentText to video · Image to videoFrom $0.024per output second
LTX-2 19BText to video · Image to video · +1From $0.014per output second
LTX-2 19B LoRAText to video · Image to videoFrom $0.018per output second

Related models and next tasks

LoRA guide

Best NSFW LoRAs for LTX-2.3

Workbench

Input

Output

Introduction

Choose a mode by starting material

Key features & highlights

Style LoRAs that take over the frame

Links from Hugging Face or Civitai

Sound on every clip

Uncensored with custom weights

Request parameters

Pricing

Use cases

Styled series

Effect shots

LoRAs your audience asks for

Prompt guide & pro tips

Use LoRAs made for LTX-2.3

Compare against no LoRA

Mind the 2 GiB limit

Use direct, public links

API quick start

One file, a whole new look

Restyle a photo as it moves

Picture and sound in one pass

More clips made with LTX-2.3 LoRA

How each mode behaves

A house style for a series

Restyle a photo as it moves

Effects and characters from the community

Uncensored community LoRAs

A glance in the manta tunnel

A gown in the projector light

A whisper in the planetarium seats

Pick a LoRA trained for LTX-2.3

Link the file, not the page

Trigger words first, then the shot

No filter from us

Tested, not guessed

Pay per result, in dollars

One key for every model

Which LoRAs work here?+

Can I use a Civitai link?+

Why did my LoRA fail?+

My LoRA loaded but nothing changed. Why?+

What strength should I use?+

How is it different from LTX 2.3 Spicy LoRA?+

Is LTX-2.3 LoRA uncensored for NSFW video?+

How is it billed?+

About LTX-2.3 LoRA

All LTX models6

Related models and next tasks

LoRA guide